论文部分内容阅读
手机用户上网时段研究与预测对手机用户行为与模式分析、网络服务内容设计、网络黏性与心理、移动互联商业智能等具有重要意义。本文结合Markov模型和关联规则模型,提出一种手机用户上网时段的混合Markov预测方法——Lift-Markov(LM)方法,并采用中国某城市4G手机用户流量上网产生的流量收费数据进行实验验证与分析。研究发现:该实验区域37.66%的手机用户个体存在明显的以天为周期的周期性特性;本文所提出的LM方法在10、20、30、40、50、60 min间隔时的平均预测准确率都优于Markov模型和Mostvalue模型,其中在60 min间隔时能达到79.75%的平均准确率,优于Markov模型(74.64%)和Mostvalue模型(64.44%);LM方法的预测准确率分布相比于其他2种模型都要窄,而且密度分布峰值最高、标准差最小,说明本文方法对人群的上网时段预测准确率较为集中与稳定,具有较好的预测性能。
Research and prediction of mobile Internet users’ time is of great significance to the behavior and pattern analysis of mobile phone users, content design of network services, network stickiness and psychology, and business intelligence of mobile internet. In this paper, Markov model and association rule model are used to propose a hybrid Markov (LM) method for cell-phone users’ access to the Internet. The proposed method is validated by using the data of traffic charges generated by 4G mobile phone users in a city in China analysis. The study found that: 37.66% of the mobile phone users in this experimental area have obvious cyclical characteristics of days; the average prediction accuracy of LM method proposed in this paper is 10,20,30,40,50,60 min Which is superior to the Markov model and Mostvalue model, with an average accuracy of 79.75% at 60 min interval, which is better than the Markov model (74.64%) and the Mostvalue model (64.44%). The prediction accuracy of the LM method is better than that of the Markov model The other two models should be narrower, with the highest density distribution and the smallest standard deviation, which shows that the proposed method is more focused and stable on the crowd prediction accuracy and has better prediction performance.